Columbia, which achieved a benchmark rating of 51.9 teraflops on 10,240 processors, is ranked second on the TOP500 List, just behind Blue Gene, IBM's supercomputer at the Department of Energy's Lawrence Livermore National Laboratory.

"Large, integrated simulation environments like those we have at Ames are crucial to NASA's missions, and Columbia has provided a breakthrough increase in our computational power," said G. Scott Hubbard, director of NASA's Ames Research Center, in California's Silicon Valley. "A high rating on the TOP500 list is an impressive achievement, but for NASA, the immediate availability to analyze important issues like 'Return to Flight' for the space shuttle, space science, Earth modeling and aerospace vehicle design for exploration, is the true measure of success."

"Columbia allows NASA to perform numerical simulations at the cutting edge of science and engineering," said Dr. Walter Brooks, chief of the NASA Advanced Supercomputing (NAS) Division at Ames. "As the largest example of an important high-end computing architecture developed in the U.S., part of this system will be available to the nation's best research teams. The swift design and deployment of Columbia has redefined the concept of supercomputer development."

With Columbia at its core, said Brooks, the NAS facility provides an integrated computing, visualization and data storage environment to help NASA meet its mission goals and the Vision for Space Exploration.

Named to honor the crew of the Space Shuttle Columbia lost Feb. 1, 2003, the new supercomputer is comprised of an integrated cluster of 20 interconnected SGI® Altix® 512-processor systems, for a total of 10,240 Intel® Itanium® 2 processors. Columbia builds upon the highly successful collaboration between NASA, Silicon Graphics Inc. (SGI) and Intel Corporation that developed the world's first 512-processor Linux server. That server, the SGI® Altix® located at Ames, was named 'Kalpana,' after Columbia astronaut and Ames' alumna Kalpana Chawla.

Within days of completion of the supercomputer's installation, Columbia achieved a Linpack benchmark rating of 42.7 teraflops on just 16 nodes with an 88 percent efficiency rating, exceeding the previously best reported performance by a significant margin. This was followed almost immediately by the 51.7 teraflop rating reported today for the entire system.

"What is most noteworthy is that we were able to post such a significant and efficient Linpack result in such a short time," said Bob Ciotti, chief systems engineer for the Columbia installation project. "Not only was the system deployed in less than 120 days, but the code used to achieve this result was conceived and developed in that same time frame, and is much more straightforward than the traditional approach. Our simplified implementation, allowed by shared memory systems like the SGI Altix, translates directly into improved effectiveness for users of our systems."

The TOP500 project was started in 1993 and provides a relatively simple way to track and detect trends in high-performance computing. Twice a year, using the Linpack benchmark as a performance measure, the project issues a list of the sites operating the 500 most powerful computer systems. The list contains a variety of information, including system specifications and major application areas.

For more information about NASA's Columbia supercomputer, visit:

http://www.nas.nasa.gov

For more information about NASA and agency programs, visit:

http://www.nasa.gov